Skip to content

newman-seed-command is an && chain and the Newman leg is disabled, so the seed has never been exercised #151

Description

@rubenvdlinde

Filed 2026-08-08 so this is not discovered by whoever flips enable-newman: true.

The state today

.github/workflows/code-quality.yml on development (c70f7b8d9):

enable-newman: false
...
newman-seed-command: 'php occ app:disable openbuild && php occ app:enable openbuild'

The Newman job is disabled, confirmed independently on #79's check rollup: quality / Integration Tests (Newman) reports SKIPPED. So the seed command has never run in CI, and the && chain has never been exercised.

Why it matters now

ConductionNL/.github#212 (issue #192) changed the shared workflow: the seed command used to run through an unquoted eval with its exit status ignored, so a failing seed exited 0 and every downstream assertion ran against an unseeded store. The status is now checked.

openbuild is reportedly the only repo in the fleet passing an && chain as its seed command. Under the fixed workflow the chain behaves as written — if occ app:disable openbuild fails, enable never runs and the chain exits non-zero, which is now surfaced. That is the correct behaviour, and it is also the first time this command's real exit status will ever be read.

Not verified here, deliberately

Whether php occ app:disable openbuild && php occ app:enable openbuild actually succeeds could only be measured by running it, and the only instance available is the shared development container that other work is using. Disabling and re-enabling an app there is disruptive to other sessions, so it was not done. The honest status is unverified, not fine.

What to do when enable-newman is flipped

  1. Read the Seed test data step's own output first. If it fails, that is the finding — an app that cannot cleanly disable/enable itself is a real problem, not CI noise.
  2. Consider whether the disable/enable round-trip is the right seed at all. The workflow comment already records two upstream blockers keeping Newman off (React\Async\await() unavailable, and SQLite-only failures in OpenRegister's MagicMapper); the second is already addressed by database: pgsql.

Cross-reference

The workflow comment says Newman stays disabled "while the openbuild#33 fixes are in flight". This issue is about the seed command shape specifically, which is independent of those.

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions